key: cord-0771654-53h0uus2
authors: Boden, M.; Cohen, N.; Froelich, J.; Hoggatt, K.; Abdel Magid, H.; Mushiana, S.
title: Mental Disorder Prevalence Among Populations Impacted by Coronavirus Pandemics: A Multilevel Meta-Analytic Study of COVID-19, MERS & SARS
date: 2020-12-22
journal: nan
DOI: 10.1101/2020.12.18.20248499
sha: 3297f46f35e1acc67987d6aa3ab5391eb542182e
doc_id: 771654
cord_uid: 53h0uus2

Coronavirus pandemics are associated with a number of well-documented threats, stressors and traumas that vary by impacted population and contribute to mental disorders. Through a systematic review and meta-analysis of research on COVID-19, severe acute respiratory syndrome (SARS) and middle east respiratory syndrome (MERS) pandemics, we investigated whether mental disorder prevalence: (a) was elevated among populations impacted by coronavirus pandemics (relative to unselected populations reported in the literature), and (b) varied by disorder (undistinguished psychiatric morbidity, anxiety, depressive, posttraumatic stress disorders [PTSD]) and impacted population (community, infected/recovered, healthcare provider, quarantined). From 60 publications (N=66,190 participants), 725 individual estimates were included in a series of multilevel meta-analyses/regressions including random effects to account for estimates nested within studies. Across disorder and population, the median summary point prevalence was 20% (95%CI=17-25%). Prevalence estimates were generally substantially higher than reported by prior research for unselected samples. Psychiatric morbidity and PTSD were most prevalent in most populations. The highest prevalence of each disorder was found among infected/recovered adults. Notably high prevalence was found for (a) psychiatric morbidity, PTSD and depression in infected/recovered adults (25-56%), (b) psychiatric morbidity and PTSD in healthcare providers (21-29%), (c) depression and PTSD in the adults in the community (15-19%), and (d) psychiatric morbidity in quarantined adults (28%). Sensitivity analyses demonstrated that overall prevalence estimates were higher for studies/estimates: (1) focused on SARS or MERS versus COVID-19, (2) conducted in Hong Kong or Korea versus other locations, obtained (3) via questionnaire versus clinician assessment and (4) with standard versus non-standard scoring, and (5) of moderate or very low versus low quality. The number of mental disorders attributable to COVID-19 will be substantial and magnitudes higher than attributable to MERS and SARS due to the vast scope and ongoing nature of the COVID-19 pandemic.

Numerous studies document the often-substantial adverse mental health impact of coronavirus pandemics (Brooks et al., 2020; Mak et al., 2009) . With the onset of COVID-19, useful information will be gained from a systematic review of this literature and quantitative summary of prevalence of mental disorders that may be common to populations impacted by coronaviruses. Through a systematic review and meta-analysis of COVID-19, severe acute respiratory syndrome (SARS) and middle east respiratory syndrome (MERS) pandemic research literatures, we addressed two questions. First, is mental disorder prevalence elevated among populations impacted by coronavirus pandemics relative to unselected populations reported in the literature? Second, does mental disorder prevalence vary by disorder (undistinguished psychiatric morbidity, anxiety, depressive, posttraumatic stress disorders [PTSD] ) and impacted population (community, infected/recovered, healthcare provider, quarantined)?

Physical exposure to a virus (e.g., through job duties), media exposure, exposure to illness and death, movement restrictions, interpersonal loss, and (for COVID-19) unemployment and economic deprivation are key pandemic-related threats, stressors and traumas that are likely to increase the risk of a mental disorder (Brooks et al., 2020; Dooley et al., 1994; Garfin et al., 2020; Lai et al., 2020; Liu et al., 2012) . Populations experiencing a greater frequency or intensity of pandemic-related stressors are likely to be at greater risk of experiencing a mental disorder (Galea et al., 2020) . Among adults infected with a coronavirus, threats to health and mortality, and disruption to routines (e.g., absence from work) will be pronounced and more impactful the greater the severity of infection (Bienvenu et al., 2018; Wu et al., 2005) . Functional impairment and disability may further increase risk for mental disorders among recovered adults (Lam et al.,

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 https://doi.org/10. .12.18.20248499 doi: medRxiv preprint 2009 Lancee et al., 2008; Lee et al., 2007; Lee et al., 2019; Mak et al., 2009) . Healthcare providers, especially front-line treatment providers who contend with threats, stressors and traumas such as repeated exposure to infected and dying people and morally ambiguous decisions regarding who receives treatment (Lai et al., 2020; Lancee et al., 2008; Liu et al., 2012) may be at increased risk of mental disorders. Quarantined adults may be at increased risk of mental disorders due to threats to health, lack of social contact, and disruptions to routine (Brooks et al., 2020) . Anxiety and depressive disorders are likely to be common as any given person may experience multiple threats and stressors that contribute to such disorders (Cisler et al., 2010; Hammen, 2005) . Infected/recovered adults and healthcare providers, in particular, may experience traumatic events (e.g., invasive treatments, witnessing death) that increase risk of post-traumatic stress disorder (PTSD; Lai et al., 2020; Mak et al., 2009 ).

Based on a systematic review of COVID-19, SARS, and MERS research, we conducted multilevel meta-analyses/regressions to derive summary prevalence estimates for multiple disorders (undistinguished psychiatric morbidity, anxiety, depressive, PTSD) in adult populations (community [including students], infected/recovered, healthcare provider, quarantined) that had been diagnosed using standardized instruments. We investigated whether mental disorder prevalence was elevated among populations impacted by coronavirus pandemics relative to unselected populations reported in the literature. For example, a meta-analysis of 157 studies from 59 countries (N ~660,000; Steel et al., 2014) found a 12-month prevalence of 15.4% for combined mood (including bipolar) and anxiety (including PTSD) disorders. Twelve-month (or less) prevalence has been found in epidemiological studies including unselected samples representative of the adult populations of: (a) China (any disorder=9.3; anxiety disorder=5.0; depressive disorder=3.6; PTSD=.2%; Huang et al., 2019) , (b) Europe (anxiety disorder=6.4, . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 ESEMeD/MHEDEA 2000 Investigators, 2004 , and (c) the United States (anxiety disorder=7.3, major depressive episode=2.2; Regier et al., 1988) . We hypothesized that, relative to these estimates, all disorders would more prevalent in all populations (we investigated) given the frequent and often impactful threats, stressors and traumas associated with coronavirus pandemics. As the frequency and impact of particular threats, stressors and traumas is likely to vary in type and by population, we further hypothesized that mental disorder prevalence would vary by disorder and impacted population.

As individual studies often provided estimates of multiple disorders from the same population, we utilized three-level meta-analytic models and included random effects to account for estimates nested within studies. The use of multilevel models allowed us to include all relevant prevalence estimates (e.g., by sex, age, income-level), thus further maximizing the information provided by any given study. Following, our study provides information not found in two published meta-analyses related to the aims of the current paper (de Pablo et al., 2020; Rogers et al., 2020) , which focused on single populations (healthcare providers, severely infected/recovered).

This study is part of a broader registered protocol (Boden, 2020) , and was conducted according to PRISMA (Moher et al., 2009) and MOOSE guidelines (Stroup et al, 2000) .

From April 15, 2020 until June 1, 2020, two study staff (MB, NC) conducted a key-word search of electronic databases PubMED, PsychINFO, Scopus, Web of Science and Google Scholar for peer-reviewed, English language publications. We searched for a broad set of studies (Boden, 2020) by forming all combinations of key words (a) avian flu, coronavirus, COVID, . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 Ebola, equine flu, flu, H1N1, Influenza, MERS, quarantine, swine flu, SARS, respiratory, crossed with (b) anxiety, depression, psychological distress, posttraumatic, emotional distress, mental health. We examined reference lists of identified publications and lists of cited studies to identify additional studies.

Studies meeting the following criteria were included in this meta-analysis: (1) available in English language, (2) peer-reviewed (no grey literature or preprints), (3) reported source data (no reviews), (4) focused on mental disorders (undifferentiated psychiatric morbidity, anxiety, depression or PTSD) related to COVID-19, MERS, SARS, (5) included quantitative estimates of mental disorder prevalence assessed by clinician diagnosis or measures with published psychometric validation data, (6) included adult participants (age >=18). Electronic database searches yielded a total of 3,930 publications that potentially met inclusion criteria. Screening of titles and abstracts yielded 270 studies that were coded (including studies of non-coronavirus outbreaks/epidemics coded as part of the broader project), with 60 studies meeting inclusion criteria (see Figure 1 ).

Publications were first coded by one of four study authors (MB, NC, SM, JF), with the majority coded by the first author. The first author developed the coding scheme and trained all other study coders through didactics, iterative feedback on coding of a subset of studies, and confirmation of codes by the first author. A subset of studies (144/270) were coded a second time to ensure consistency of coding. Disagreements between primary and secondary codes were resolved by the first author.

Coded study attributes include mental disorder prevalence in terms of: (a) sample size . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. We coded all relevant estimates provided in each study, including estimates for different disorders and the same disorder assessed by different measures, at different time-points, or among different populations and subpopulations (e.g., quarantined versus community, males versus females, doctors versus nurses). Thus, we coded sociodemographic and socioeconomic characteristics but did not include them in moderator analyses due to inconsistent reporting across studies. We also coded study attributes not included in the reported analyses (i.e., means and standard deviation for continuous measures of mental disorders/symptoms).

We manually calculated size of sample positive for a disorder when not reported by the study. For questionnaire measures, we coded size of sample positive for a disorder consistent with the study authors, who typically utilized established cutoffs reported in prior research.

Estimates obtained from non-standard cutoffs were identified for sensitivity analysis. When severity thresholds (only) were provided (e.g., mild, moderate, severe), participants in moderate or higher categories were coded as positive for a disorder unless otherwise specified in prior research.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 Study quality was assessed by the Systematic Assessment of Quality in Observation Research (SAQOR) tool (Ross et al., 2011) with several changes to increase the applicability of the system to our coding of prevalence (see Supplement B).

Using the metafor (Viechtbauer, 2010) package in R, we conducted a series of mixedeffect multilevel meta-analyses/regressions. Our main analyses consisted of four metaanalyses/regressions, each of which included a random effect for studies and estimates within studies. First, we conducted a meta-analysis to obtain an overall summary prevalence estimate across disorders and populations. Second, we conducted a meta-regression including (dummycoded) disorder as a fixed effect moderator to obtain summary prevalence of each disorder across populations. We conducted a third meta-regression including (dummy-coded) population to obtain overall mental disorder prevalence in each population. The fourth meta-regression, including dummy-coded disorder, population, and their product, provided summary estimates of individual disorders in individual populations. Studies that utilized the same samples were analyzed as a single study.

Models were implemented using restricted maximum-likelihood estimation. Study prevalence estimates (i.e., sample size with disorder divided by total sample size) and 95% confidence intervals (CI) were Freeman-Tukey double arcsine transformed (Freeman et al., 1950) for improved statistical properties. Fixed effects and CI were back-transformed (Miller, 1978) to provide summary prevalence estimates. The I 2 statistic measured percentage of heterogeneity due to true between-study and between-estimate differences, and the interclass correlation coefficient (ICC), the association between underlying estimates. Profile plots indicated whether all variance components were statistically identifiable.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10. 1101 /2020 As part of sensitivity/quality analyses, we re-conducted our four main analyses using a cluster-robust estimator of standard errors (similar to the Eicker-Huber-White method) useful when multiple, dependent outcomes come from individual studies (Hedges et al., 2010) .

Additionally, in a series of meta-regressions, we examined whether the overall summary effect varied by: (1) pandemic, (2) location, (3) timing, (4) measure (self-report questionnaire versus clinician assessment), (5) scoring (non-standard/unknown versus standard scoring of measure), and (6) study quality (very low, low, moderate). Analysis of publication bias was not conducted as most studies did not use inferential statistics to test hypotheses regarding prevalence, and thus, were not more likely to be published because of statistically significant results.

From 60 publications (of which six studies/three-pairs utilized the same or partially overlapping samples) including 66,190 participants, 725 individual estimates were obtained. In main analyses, approximately 99% of variance between-and within-studies was due to true variation rather than sampling error. As expected, between-study heterogeneity was larger (Range I 2 = 72.1-73.3%) than within-study heterogeneity (Range I 2 = 25.3-26.8%). ICC ranged from .73 to .74, indicating a strong association between estimates within studies. Models with random effects fit significantly better than respective models without (all p<.001). Profile plots indicated all variance components were statistically identifiable. Summary estimates for individual disorders/populations were based on varying numbers of studies, individual effects and participants (see Table 1 ). CI were large for summary estimates derived from few studies/estimates (e.g., quarantined), thus indicating greater imprecision.

The median summary point prevalence for mental disorders across disorders and populations was 20% (95%CI=17-25%). Prevalence estimates of 20% or higher were found for . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint 11 (44%) disorders/populations with a high of 56% for psychiatric morbidity among infected/recovered adults (see Table 1 ). Occurring in 32% (95%CI=27-37%) of the overall sample, psychiatric morbidity had the highest prevalence across all populations, and in each population other than depression in community adults (see Figure 1 ). Across all populations, PTSD had the second highest prevalence (21%, 95%CI=17-25%), followed by depression (17%, 95%CI=14-22%) and anxiety (12%, 95%CI=8-15%).

Across disorders, the highest prevalence was found for infected/recovered adults (30%, 95%CI=21-40%), followed by healthcare providers (20%, 95%CI=15-26%), community adults (16%, 95%CI=11-23%), and quarantined adults (12%, 95%CI=4-22%). The highest prevalence of each disorder was found for infected/recovered adults, with 25-56% positive for depression, PTSD or psychiatric morbidity. Healthcare providers had high psychiatric morbidity prevalence (29%, 95%CI=23-36%) and PTSD (21%, 95%CI=16-27%) relative to other disorders. Adults in the community had high prevalence of depression (19%, 95%CI=12-26%) and PTSD (15%, 95%CI=9-23%) relative to other disorders. Quarantined adults had much higher prevalence of psychiatric morbidity (28%, 95%CI=5-59%) versus other disorders, but summary prevalence based on few studies/individual estimates were imprecisely estimated.

In sensitivity analyses, use of the robust estimator resulted in wider confidence intervals on average for summary estimates (Mean Difference=.03, Range= -.62 to .18) and backtransformed point prevalence estimates (Mean Difference=2%, Range= -54% to 13%; see Figure   2 and Supplement C). Most studies (a) focused on SARS, (b) were conducted in China and (separately) Hong Kong, included estimates (c) measured during the acute phase and obtained (d) via questionnaire and (e) with standard scoring, and (f) were of low or very low quality (see Table 2 ). The overall prevalence estimates across disorders and populations were higher for . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint 

Across 60 individual studies, one in five adults in populations impacted by a coronavirus pandemic were demonstrated to or reported signs or symptoms indicative of a mental disorder diagnosis. Our summary estimates were higher, often substantially than those found in a metaanalysis of 157 studies from 59 countries (Steel et al., 2014) and in epidemiological studies of the adult populations of China (Huang et al., 2019) , Europe (ESEMeD/MHEDEA 2000 Investigators, 2004 , and the United States (Regier et al., 1988) . Undifferentiated psychiatric morbidity, the most prevalent disorder in most of our study populations, had a much higher prevalence than in the prior studies referenced above, as did anxiety and depressive disorders in all but quarantined populations. PTSD, the second most prevalent disorder in all study populations, had a much higher prevalence than reported for unselected Chinese adults (Huang et al., 2019) , and was comparable to lifetime prevalence of PTSD among combat veterans (Weiss et al., 1992) . Our results further demonstrate that coronavirus pandemics are associated with a substantial adverse mental health impact as manifested by elevated rates of multiple mental disorders in several populations. Our search strategy and analytical methods allowed us to utilize all relevant data from individual studies, while accounting for dependencies in those data.

Furthermore, our focus on peer-reviewed studies utilizing standardized measures strengthens our confidence in the validity of our results.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10. 1101 /2020 Infected or recovered adults had the highest prevalence of all disorders and each individual disorder, with notably higher prevalence relative to unselected adults in prior studies (ESEMeD/MHEDEA 2000 Investigators, 2004 Huang et al., 2019; Regier et al., 1988) .

Prevalence of PTSD was quite high (28%) and consistent with research finding PTSD is common among mechanically ventilated and intensive care unit patients (Bienvenu et al., 2018; Griffiths et al., 2007) . Severe infection appears to be a potent risk factor for mental disorders, and may contribute to functional impairment and disability that continues to adversely impact the mental health of recovered patients (Mak et al., 2009) . Even minor infections may be quite threatening and stressful in people with pre-existing physical or mental health conditions. Nuanced investigations of links between mental disorders and infection severity, symptom presentation and medical treatments, and facets of (dys)function post recovery will be useful for mental health intervention planning. Immediate research of this type is needed for COVID-19.

Demonstrating the substantial threat, stress and trauma faced by healthcare providers during a coronavirus pandemic, high mental disorder prevalence was found among healthcare providers relative to unselected populations in prior studies (ESEMeD/MHEDEA 2000 Investigators, 2004 Huang et al., 2019; Regier et al., 1988) . Notably high PTSD prevalence may reflect exposure to numerous potentially traumatic events (e.g., suffering and death of patients, morally ambiguous decisions and moral injury). Job stresses (e.g., long and inflexible hours, fluid work environments and job duties) may contribute to high prevalence of depression, along with burnout.

Even adults in the community had elevated prevalence of mental disorders relative to unselected populations in prior studies. The high prevalence of depression, in particular, might reflect the stress of living and coping with the threat of infection, societal dysfunction, and/or . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint limited availability of routines and resources. Countries that continue to be impacted by COVID-19 are likely to experience increased incidence of depression in the general population over time.

Findings of lower prevalence of most disorders among quarantined populations must be interpreted cautiously given the small number of studies/estimates from which these summary estimates were derived (e.g., 58 [83%] of individual estimates came from one of the five studies providing estimates for quarantined adults).

Several issues should be considered when interpreting these results. First, as indicated by large CI, some studies produced imprecise summary prevalence estimates, likely due to small sample sizes. Second, most studies were of low or very low quality. Summary estimates varied somewhat by study quality, though summary estimates were identical for moderate and very low quality studies. Third, the true prevalence of diagnosable mental disorders in populations impacted by coronavirus epidemics is likely to be lower than the magnitude of summary estimates, derived mostly from studies using self-report questionnaires that do not precisely assess mental disorder diagnostic criteria. Only a subset of included studies utilized goldstandard clinical interviews (e.g., Structured Clinical Interview for DSM) and assessment instruments specifically designed to screen for mental disorders (e.g., Kessler et al., 2002) .

Indeed, sensitivity analyses revealed higher overall summary prevalence obtained from questionnaires versus clinician assessment. Somewhat mitigating this concern: (a) almost all studies utilizing questionnaires implemented empirically validated cutoffs to identify participants with mental disorders (e.g., Plummer et al., 2016); and (b) studies that did not implement standard questionnaire cutoffs (and those that utilized unknown scoring algorithms to identify mental disorders) yielded similar overall summary prevalence estimates as those that did.

As applied to understanding and planning for COVID-19 (Boden et al., 2020) , summary . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint estimates must be interpreted cautiously. Overall prevalence of all disorders across all populations was lower for COVID-19 versus MERS and SARS. All COVID-19 studies focused on mental disorder prevalence during the acute phase of the pandemic, whereas MERS and SARS studies included acute and longer time-frames. Yet, the overall mental disorder prevalence varied little between acute and longer time-frames. It will be important to investigate whether mental disorder prevalence increases in countries with prolonged periods of SARS-CoV-2 infection and economic hardship (e.g., United States, Brazil). We hypothesize that such countries, and areas within these countries may experience increased rates of the disorders examined in this study, in addition to disorders of despair (e.g., alcohol dependence; Petterson et al., 2020) and related to disease threat (e.g., illness anxiety disorder, somatic symptom disorder).

Regardless, the raw number of mental disorders attributable to COVID-19 will be substantial and magnitudes higher than attributable to MERS and SARS due to the vast scope and ongoing nature of the pandemic.

Preliminary evidence suggests that COVID-19 exacerbates existing inequalities that harm racial/ethnic minorities, economically disadvantaged people, people with 'essential' job types, etc. (Hooper et al., 2020; van Dorn et al., 2020) . Additional research is needed to quantify the mental health impact of COVID-19 in these subpopulations, and to document disparities in assessment and treatment of COVID-19-related mental disorders. Future research that examines the potentially manifold pathways to individual outcomes among subpopulations most at-risk will be instrumental in intervening in a cost-effective, effective and equitable manner. Future research that utilizes rigorous and validated subject recruitment and mental disorder assessment methods can improve upon the low quality of the majority of studies included in our study.

Additional limitations can be addressed through assessment of pre-existing mental disorders . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint among study samples, rigorous assessment of time-periods at risk, and/or inclusion of control conditions (e.g., people exposed to but not infected by COVID-19 as a comparator for infected/recovered people). Studies that do so will more directly measure the impact of COVID-19-related threats, stressors and traumas on mental disorder prevalence, incidence, and rate ratios, which will be essential to intervention planning and implementation.

Our study provides data useful for understanding and potentially, intervening to alleviate the mental health impact of COVID-19 and future coronavirus (and other) pandemics. A datadriven approach will facilitate resource allocation to provide effective (and cost-effective) mental health interventions for people with the greatest need during and following a coronavirus pandemic. For example, the high prevalence of PTSD among infected/recovered people indicates a need for effective evidence-based treatments for PTSD as part of convalescence (e.g., prolonged exposure for PTSD; Powers et al., 2010), and providers trained to administer these interventions. A data-drive approach to resource allocation will be especially useful to countries, states/provinces and healthcare systems for which treatment need is expected to exceed existing mental health resources.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 outbreak on health care workers in a medium size regional general hospital in Singapore. . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint associated with mental health outcomes among health care workers exposed to . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 impact of the COVID-19 outbreak: an early report on the Italian general population. Brain, Behavior & Immunity. 2020; 87:75-79. doi:10.1016 /j.bbi.2020 Moher D, Liberati A, Tetzlaff J, Altman DG, Prisma Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Medicine. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint . CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101/2020.12.18.20248499 doi: medRxiv preprint Figure Caption. Summary point prevalence estimates and 95% confidence intervals from multilevel meta-analyses with a robust standard error estimator.

. CC-BY-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted December 22, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 

Middle East Respiratory Syndrome-Corona Virus (MERS-CoV) associated stress among medical students at a university teaching hospital in Saudi Arabia

Survey of stress reactions among health care workers involved with the SARS outbreak

Psychiatric symptoms after acute respiratory distress syndrome: a 5-year longitudinal study

Estimating the mental health impact of COVID19 on United States Populations

The psychological impact of quarantine and how to reduce it: rapid review of the evidence

Depression and unemployment: panel findings from the Epidemiologic Catchment Area study

Prevalence of mental disorders in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project

Transformations related to the angular and the square root

The mental health consequences of COVID-19 and physical distancing: The need for prevention and early intervention

The novel coronavirus (COVID-2019) outbreak: Amplification of public health consequences by media exposure

Grading quality of evidence and strength of recommendations

The prevalence of post traumatic stress disorder in survivors of ICU treatment: a systematic review

Stress and depression

Do psychiatric patients experience more psychiatric symptoms during COVID-19 pandemic and lockdown? A case-control study with service and research implications for immunopsychiatry

SARS control and psychological effects of quarantine

Robust variance estimation in meta-regression with dependent effect size estimates

Posttraumatic stress disorder in convalescent severe acute respiratory syndrome patients: a 4-year follow-up study

COVID-19 and racial/ethnic disparities

Prevalence of mental disorders in China: a cross-sectional epidemiological study. The Lancet Psychiatry

Generalized anxiety disorder, depressive symptoms and sleep quality during COVID-19 outbreak in China: a web-based cross-sectional survey

prolonged exposure for posttraumatic stress disorder

One-month prevalence of mental disorders in the United States: Based on five epidemiologic catchment area sites

Understanding, compliance and psychological impact of the SARS quarantine experience

Psychiatric and neuropsychiatric presentations associated with severe coronavirus infections: a systematic review and meta-analysis with comparison to the COVID-19 pandemic. The Lancet Psychiatry

Quality assessment of observational studies in psychiatry: an example from perinatal psychiatric research

The effects of disease severity, use of corticosteroids and social factors on neuropsychiatric complaints in severe acute respiratory syndrome (SARS) patients at acute and convalescent phases

Severe acute respiratory syndrome-related psychiatric and posttraumatic morbidities and coping responses in medical staff within a primary health care setting in Singapore

Psychological impact of the SARS outbreak on a Singaporean rehabilitation department

Hospital workers' psychological resilience after the 2015 Middle East respiratory syndrome outbreak

The global prevalence of common mental disorders: a systematic review and meta-analysis 1980-2013

MOOSE guidelines for metaanalyses and systematic reviews of observational studies

Prevalence of psychiatric morbidity and psychological adaptation of the nurses in a structured SARS caring unit during outbreak: a prospective and periodic assessment study in Taiwan

Journal of Traumatic Stress: Official Publication of The International Society for Traumatic Stress Studies

The psychological impact of the SARS epidemic on hospital employees in China: exposure, risk perception, and altruistic acceptance of risk

Alcohol abuse/dependence symptoms among hospital employees exposed to a SARS outbreak

The psychological burden experienced by Hong Kong midlife women during the SARS epidemic. Stress and health

Mental health and psychosocial problems of medical health workers during the COVID-19 epidemic in China